Unblu

概要

The unblu-mcp project is a Model Context Protocol (MCP) server specifically designed for seamless integration with Unblu deployments. It empowers AI assistants by providing secure and token-efficient access to Unblu's extensive API, featuring over 300 endpoints. By implementing progressive disclosure, the server significantly reduces token usage for large language models, offering clear tool interfaces, smart discovery capabilities, and robust safety mechanisms, including client confirmations and optional server-side policy enforcement. It supports the latest MCP specification and integrates advanced features like response caching and flexible connection providers.

主な機能

• Progressive Disclosure: Uses 5 discovery tools to represent 300+ API operations, drastically cutting token usage for AI.
• Token-Efficient Access: Includes field filtering and response truncation to minimize data transfer and LLM token consumption.
• Robust Safety Controls: Features `destructiveHint` for client confirmations and optional server-side policy enforcement via Eunomia.
• 1 GitHub stars
• Built with FastMCP 2.14+: Leverages advanced features like MCP annotations, response caching, and policy-based authorization.
• Full API Coverage: Provides access to all Unblu REST API v4 endpoints.

ユースケース

• Developing AI assistants that can interact with and manage Unblu conversations, users, and bots.
• Integrating large language models securely into Unblu environments with fine-grained control over API access.
• Optimizing token consumption for AI agents needing to perform operations against Unblu's comprehensive API.